Measuring heart rate with your phone camera: how it works, and how accurate it really is

When there's no Apple Watch or chest strap around, the camera on your phone can still give you a rough sense of your pulse. Hold a fingertip over the rear lens, or point the front camera at your face, and an app can turn the flicker of light and color it picks up into a heart rate reading in under a minute. This is informational material, not medical advice — what a phone camera produces is an estimate, not a measurement, and it isn't a diagnosis of anything. The honest question this article tries to answer is the one that matters most before you trust a number like that: how close is it, really, to what a proper heart rate device would tell you?

How it actually works

The technique behind camera heart rate has a name — photoplethysmography, or PPG — and it predates the smartphone by decades. PPG is a simple, inexpensive optical method that tracks changes in blood volume within a bed of microvessels in living tissue: with every heartbeat, blood volume in the tissue rises and falls, and the light reflected or transmitted through that tissue rises and falls with it. Physiological-measurement researchers describe the resulting signal as having two parts — a slowly drifting "DC" baseline shaped by breathing, sympathetic nervous activity, and thermoregulation, and, riding on top of it, a pulsing "AC" component that tracks each individual heartbeat (Allen, 2007). A phone camera does the same job with different hardware: instead of a dedicated LED-and-photodiode sensor, it watches for the same faint, frame-to-frame shifts in skin color and brightness — whether that's a fingertip pressed against the rear lens or a face held in front of the front camera — and the period of that AC pulsation is what the app reports back as a heart rate.

How close is it to your real heart rate

Camera-based PPG isn't an idea that was validated once and left alone — it has been checked directly against clinical references more than once, and the most useful of those checks tested real consumer apps rather than a lab prototype. A 2017 study put four smartphone heart rate apps through their paces on 108 patients, each of whom also had a simultaneous ECG and pulse oximeter reading as reference. The pulse oximeter itself, the more directly comparable reference, came in with a mean absolute error of about 2 ± 0.35 beats per minute against ECG — a useful yardstick for what "good" looks like here. Against that yardstick, the contact apps, which use a fingertip pressed on the rear camera, did reasonably well: Heart Fitness averaged 2 ± 0.5 bpm off, Instant Heart Rate 4.5 ± 1.1 bpm off. The contactless apps, which read a face in front of the front camera, did noticeably worse: Whats My Heart Rate averaged 7.1 ± 1.4 bpm off, Cardiio 8.1 ± 1.4 bpm off (Coppetti et al., 2017). The authors' own conclusion is worth repeating plainly: the contact, finger-based apps had both higher usability and better accuracy than the contactless, face-based ones. The honest summary of these numbers is that even the best camera-based estimate lands a few beats per minute off a clinical device — close enough to be a useful gut-check, not close enough to trust the way you'd trust an ECG. If you have to choose between the two methods, the finger is the more reliable one.

What throws off the estimate

Because camera heart rate is reading a faint color signal off your skin, anything that weakens or muddies that signal drags the estimate further from the truth. A 2021 study that tested face-based readings systematically found the estimate was most accurate when the light in front of the face exceeded 500 lux — tested at 100, 300, 500, and 700 lux — and got worse in dimmer rooms. Motion mattered even more: the average error in the interval between heartbeats nearly doubled while participants talked (53.9 ms) compared with holding still (23.9 ms), and a related measure of heart-rate variability that correlated almost perfectly with a reference device while sitting still (0.976) collapsed to essentially no correlation while talking (0.047). The same study found that a distance of about 0.5 meters and a frame rate of 30 fps or more were enough, so the practical takeaway is narrower than it sounds: hold still, face good front light, and keep the phone steady, and the reading has a real chance of being close (Tohma et al., 2021).

There's a second, less obvious source of error, and it doesn't go away with better lighting or a steadier hand: skin tone. Camera-based PPG depends on light reflecting back through the skin, and melanin absorbs more of that light — so, all else equal, the signal it produces is weaker for darker skin. A 2025 audit of roughly 100 rPPG studies and their public datasets found that darker skin tones (Monk scale 4–10) made up under 25% of the people represented, against roughly 45% for lighter skin tones (Monk scale 1–3). The authors put the consequence plainly: "Datasets characterized by underrepresentation of darker skin tones or broad ethnic diversity limit the reliability and fairness of rPPG algorithms in real-world applications" (Bondarenko et al., 2025). That's not a footnote — it means the accuracy figures above, measured on whoever happened to be in a given study's sample, may not hold evenly across everyone reading this.

What this isn't

It's worth being blunt about the boundaries here, because it's easy for "camera heart rate" to sound like more than it is. It estimates one thing — pulse rate — and nothing else: not blood pressure, not blood oxygen (SpO₂), not any other cardiac metric, no matter how the light on your face flickers. It is not a substitute for a pulse oximeter or an ECG, and it isn't built or validated for diagnosing anything. What it is: a convenient, rough estimate of your pulse for the moment when your watch isn't on your wrist — nothing more, nothing less.

Where this fits in MeteoHealth

MeteoHealth uses this same principle — camera-based PPG, by fingertip or by face — as a way to log a heart rate estimate on the days your Apple Watch isn't around, so a gap in the day doesn't have to be a gap in your data. That estimate then joins the same on-device engine that looks at your Apple Health data and local weather, the same way every other measure in this series does — it's compared, not diagnosed, and it stays an estimate at every step, never promoted to something more precise than the method behind it actually is. Nothing about the reading leaves your device to make that comparison.

Bottom line

Camera-based heart rate works on real, well-understood optics — the same photoplethysmography principle used in clinical pulse oximeters — and in good conditions, a fingertip against the rear camera lands within a few beats per minute of a proper ECG. But "good conditions" is doing real work in that sentence: enough front light, a still hand and a still face, and — less within your control — a camera-and-dataset ecosystem that has historically read lighter skin tones more reliably than darker ones. None of that makes the number useless; it makes it exactly what it is — a convenient estimate for a moment without a watch, not a measurement to lean on.